1. Field of the Invention
The present invention relates to an object information processing apparatus for processing input image data to obtain information about an object contained in the image, an image processing system having a function of recognizing the object contained in the image using the object information, and a game apparatus and an image processing method using the image processing system.
2. Description of the Related Art
As high-performance and low-cost image pickup devices (e.g., CCDs, CMOS imagers, etc.), image processing DSP devices, and the like have been developed, a number of products have incorporated a camera module. Recently, camera modules are often employed even in mobile devices (e.g., cellular telephones, etc.), toys, peripheral devices thereof, and the like.
Camera modules are utilized to input image data. Such an image data input method can handle a large amount of information simultaneously and therefore is considerably efficient compared to data input methods which employ a key board, a jog dial, a joystick or the like to input operating commands.
However, in conventional image data input methods using a camera module, input image data is typically displayed as an input image on a monitor screen, and subsequently a final process, i.e., various judgment (store or send data, etc.) is performed by a human.
In some game apparatus or the like, the user performs actual movements in order to enhance virtual reality. To this end, a number of techniques for sensing movements have been developed. Particularly, if the movements of the user can be input as image data into an image processing system so that the movements can be recognized, this technique may be useful for game apparatuses.
After the movements of the user have been input as image data to an image processing system, the central processing unit (CPU) of a computer needs to be used to eventually recognize the movements of the user. Before the eventual recognition, it is necessary to extract required information from the image data. Conventionally, a number of techniques for extracting an object from image data have been proposed.
For example, an object recognition apparatus for recognizing a moving object has been proposed in Japanese Laid-Open Publication No. 6-266841. In this technique, a stationary object is detected in image data by feature extraction; background data is obtained; and the background data is subtracted from current image data to recognize a moving object.
Japanese Laid-Open Publication No. 63-88684 discloses an object recognition method, in which a luminosity gradient vector is calculated based on the luminosity concentrations of adjacent pixels, and a difference in angle of the luminosity gradient vector is used to recognize an object.
Japanese Laid-Open Publication No. 59-132075 discloses an image processing apparatus, in which only the contour of an object is stored in an image memory before recognition is performed by a central processing unit in order to reduce a processing time required for recognition.
The above-described conventional object recognition apparatus (Japanese Laid-Open Publication No. 6-266841) requires a memory for storing a large amount of data for each step of object feature extraction, stationary object recognition, background data calculation, image subtraction, and moving object recognition. These steps require a large computation amount. Moreover, if real-time processing is required for the step, a high-performance CPU or processing operation circuit needs to be employed.
Also in the conventional object recognition method (Japanese Laid-Open Publication No. 63-88684), image data needs to be stored in order to compare image data between adjacent pixels during processing operations, and therefore, an image memory for storing at least one frame of image data is required.
Also in the conventional image processing apparatus (Japanese Laid-Open Publication No. 59-132075), data representing the contour of an object is stored in a memory before object recognition is performed by a CPU, and therefore, it is not easy to perform object recognition in real time. Particularly, real-time object recognition cannot be achieved by a CPU having poor processing performance or a one-chip microcomputer incorporating the CPU are used. In the image processing apparatus, an image memory for storing data representing the contour of an object is required, so that it is not easy to reduce cost.
In the above-described object recognition techniques, sophisticated processing operations are performed using an image memory to carry out complicated extraction for object recognition. Therefore, a memory for storing a large amount of image data for image processing and a CPU or processing operation circuit for processing the data at high speed are required, so that it is not easy to achieve an image processing system at low cost. Therefore, use of image data for data input is not suitable for inexpensive products, such as toys or the like, although it is effective as a data input method.
The variety and quantity of object extraction information obtained by conventional object recognition techniques are large, information required for object recognition varies depending on the image processing system.